ABSTRACT
My contribution to this historic celebration of the COPSS concerns the field of survey sampling, its history and development since the seminal paper by Neyman (1934), current orthodoxy, and a possible direction for the future. Many encounter survey sampling through the dull prism of moment calculations, but I have always found the subject fascinating. In my first sampling course, I remember being puzzled by the different forms of weighting in regression — by the inverse of the probability of selection, or by the inverse of the residual variance (Brewer and Mellor, 1973). If they were different, which was right? My early practical exposure was at the World Fertility Survey, where I learnt some real-world statistics, and where the sampling guru was one of the giants in the field, Leslie Kish (Kish et al., 1976). Kish was proud that the developing countries in the project were more advanced than developed countries in publishing appropriate estimates of standard error that incorporated the sample design. Always engaging, he shared my love of western classical music and tolerated my model-based views. More recently, I spent time helping to set up a research directorate at the US Census Bureau, an agency that was at the forefront of advances in applied sampling under the leadership of Maurice Hansen.
